具有特征矩阵的 Stl 迭代器

Question

I have some functions that involves stl iterators and works with types like std::vector<Eigen::Vector2d> . For example:

template<typename T>
bool isLeftOf(const Eigen::Vector2<T>& a,
              const Eigen::Vector2<T>& b) {
  return (a.x() < b.x() || (a.x() == b.x() && a.y() < b.y()));
}

int main()
{
  std::vector<Eigen::Vector2i> myVec;
  myVec.push_back(Eigen::Vector2i::Random(2));
  myVec.push_back(Eigen::Vector2i::Random(2));
  myVec.push_back(Eigen::Vector2i::Random(2));
  myVec.push_back(Eigen::Vector2i::Random(2));
  myVec.push_back(Eigen::Vector2i::Random(2));
  
  Eigen::Vector2i element = *std::min_element(myVec.begin(), myVec.end(), isLeftOf<int>);

  return 0;
}

As you can see I create std::vector<Eigen::Vector2i> myVec and work with Eigen::Vector2<T> in function isLeftOf when calling std::min_element .

Now I'm experincing some troubles while working with std::vector<SomeEigenType> and I'm looking for a way to work with the same isLeftOf(Eigen::Vector2...) and stl functions but I don't understand how.

In Eigen documentation or in the forum there is some information how to perform stl operations on Eigen::Vector or Matrix but they work with plain numbers of matrix so I cant send Eigen::Vector2 to my isLeftOf function.

Is there a way to use stl functions with Eigen::Matrix and process condition in my functions like isLeftOf that accept Eigen::Vector types?

Answer 1

Using the master version of Eigen, you can use STL-like iterators which access rows or columns of matrices:

Assuming myVec is a matrix you can write this to get the "left-most" column:

  Eigen::Vector2i element = *std::min_element(
      myVec.colwise().begin(), myVec.colwise().end(), 
      [](auto const& a, auto const& b){
          return std::lexicographical_compare(a.begin(), a.end(), b.begin(), b.end());
      });

Godbolt-Demo: https://godbolt.org/z/n1Y8hf

Answer 2

I have not found the best solution but for now I decided to use lambda functions:

#include <algorithm>
#include <Eigen/Dense>

int main()
{
  Eigen::MatrixX2f M;

  /*
   *  fill matrix M as you need
   * */
  
  // initialize original index locations from 0 to N-1
  Eigen::VectorX<Eigen::Index> idx =
      Eigen::ArrayX<Eigen::Index>::LinSpaced(
        M.rows(), 0, M.rows()-1);

  std::function<bool(const Eigen::Index &, const Eigen::Index &)> isLeftOf_fun = 
      [&M](
      const Eigen::Index& row1,
      const Eigen::Index& row2)->bool
  {
    return (M(row1,0) < M(row2,0) || (M(row1,0) == M(row2,0) && M(row1,1) < M(row2,1)));
  };

  Eigen::Index ia = *min_element(idx.begin(), idx.end(), isLeftOf_fun);
  Eigen::Vector2f element = M.row(ia);

  return 0;
}

The main idea is to apply stl iterator on index vector wich has values from 0 to N-1 (where N is the number of rows of matrix) and process the chosen rows in lambda (or in standard) function.